How do I find the period or frequency of a signal?

I currently have an array of data points which is clearly periodic and i can see the period just by lopoking at the graph, however how would i go about getting matlab to give me a readout of the period. One problem with finding the distance between peaks is that they are faily noisey so was thinking of finding distance between the dips, but i am having trouble workiing out how i would go about doing that.

Answer 1

[Star Strider]

I would first detrend (link) your data, then use findpeaks (link) to identify the peaks and their locations. Use the location information to calculate the period and from that, frequency.

[Phillip C]

Thanks for your help. I have used detrend to remove the falling trend of my signal (g) and then found the peaks which have a prominence of at least a quarter of the signal and found their average difference,

h=detrend(g);
[p,peaks]=findpeaks(h,'MinPeakProminence',(max(h)-min(h))/4);
mean(diff(peaks))

[Star Strider]

As always, my pleasure.

Note that if you have a specific time vector, for example ‘tv’, using your findpeaks call, you need to determine the period as:

Period = mean(diff(tv(peaks)));

That will give you the period in terms of your sampling times.

If my Answer helped you solve your problem, please Accept it!

[Xu]

Isn't peak wave period and wave period are two different things?

[jb4earth]

@Xu you are correct

PEAK period should be calculated as:

peak_wave_period = max(diff(your_time_vector(peaks)))

[DoubleO7]

I was wondering if anyone could expand upon this answer. What do you mean by time vector? Would an example of a time vector be:

tv = 1:100;

Is writing the following code correct for displaying the period?

[pks,locs] = findpeaks(data,tv)
period = max(diff(locs))

I appreciate your input.

Answer 2

[Dmitry Grigoryev]

Start by substracting the sliding average from the raw signal.

Then shift the signal by rougly 1 period and use autocorrelation to find the exact period value.